3D counting and tracking

Extends standard 2D workflows to analyze volumetric (3D) image data typically acquired by confocal scanning or a spinning disk. Light-sheet imaging is becoming common too.

Instead of analyzing individual slices, these workflows process complete Z-stacks as single volumes, detecting and measuring true 3D objects. Although 3D imaging and analysis is much closer to the reality it brings several challenges.

Overview of 3D analysis

Key differences from 2D

Dimensionality:

• 2D: Objects are patches of connected pixels (X, Y)
• 3D: Objects are volumes of connected voxels (X, Y, Z)

Measurements:

• 2D: Area, perimeter, circularity (µm², µm)
• 3D: Volume, surface area, sphericity (µm³, µm²)

Processing:

• 2D: Frame-by-frame through time-lapse
• 3D: Volume-by-volume through time series

Visualization:

• 2D: Single slice or maximum projection
• 3D: Volume rendering, Z-projections

When to use 3D analysis

Required for:

• Spherical or irregular 3D structures (organoids, spheroids, cell clusters)
• True volumetric measurements
• 3D spatial relationships between objects
• Tracking particles moving in 3D space

Not necessary for:

• Flat monolayer cell cultures
• Objects confined to single focal plane
• When 2D measurements are sufficient (e.g., cell counting)

Common 3D workflow structure

All three workflows share these 3D-specific components:

3D segmentation nodes

Replace 2D segmentation with volumetric equivalents:

• Bright Spots 3D instead of Bright Spots
• Threshold 3D instead of Threshold
• Cellpose 3D instead of Cellpose

These nodes detect objects across the entire Z-stack simultaneously.

3D measurement nodes

Provide volumetric features:

• Object 3D Count - counts objects per volume
• Object 3D Measurement - measures 3D morphology
• Time and Center 3D - extracts 3D centroids

Processing mode

GA3 automatically switches to volume-by-volume processing when 3D nodes are present. Instead of processing frame 1, frame 2, etc., it processes volume 1, volume 2, etc.

Calibration requirements

3D analysis requires proper Z-calibration:

• XY calibration: Microns per pixel (typically 0.1-1 µm/pixel)
• Z calibration: Microns per slice (typically 0.2-2 µm/slice)

Incorrect Z-calibration produces distorted measurements (spheres appear as ellipsoids).

Workflow 1: 3D object counting

Detects and counts 3D objects across time-series volumes, measuring object counts and intensities per volume.

  
---
config:
  look: handDrawn
  theme: neutral
---
graph TD
  subgraph "Object features"
    meas3d[3D Object Measurement]
    table1[Table - Current Volume]
    histo[Histogram]
  end

  subgraph "Object count"
    count3d[3D Object Count]
    accum[Accumulate Records]
    table2[Table - All Volumes]
    chart[Line Chart]
  end

  subgraph "Presentation"
    layout1[Stacked Layout - Left]
    layout2[Stacked Layout - Right]
    display[Display]
  end

  subgraph Outputs
    savechn[Save Channels]
    savebin[Save Binaries]
    savetab[Save Tables]
  end

  %% Main workflow
  channels[Channels] -->|Cyan| spots3d[Bright Spots 3D] -->|Bin| count3d
  channels -->|Fluo| count3d
  spots3d -->|Bin| meas3d
  channels -->|All| meas3d

  count3d -->|Records| accum
  accum -->|Records| table2
  accum -->|Records| chart

  meas3d -->|Records| table1
  meas3d -->|Records| histo

  %% Presentation
  table1 --> layout1
  table2 --> layout2
  histo --> layout2
  chart --> layout2

  layout1 --> display
  layout2 --> display

  %% Outputs
  channels -->|All| savechn
  spots3d -->|Bin| savebin
  display -->|Results| savetab

Bright Spots 3D detection

Bright Spots 3D detects fluorescent spots as 3D objects:

Parameters:

• Diameter: 8 microns - expected object diameter in XY
• Contrast: 100 - minimum brightness above background
• Z-Axis elongation: 1:2 - typical anisotropy
• Output type: “Centers” - creates centroids for fast rendering but only for counting

The algorithm searches for local intensity maxima and grows them in 3D until background is reached or growth limits are hit.

Object 3D Count measurement

Object 3D Count provides per-volume statistics:

Measured features:

• Time: Timestamp of the volume
• Object Count 3D: Total number of 3D objects in volume
• Mean Intensity: Average intensity per channel (aggregated across objects)

This produces one row per volume (time point), summarizing the entire 3D field.

Object 3D Measurement

Object 3D Measurement measures individual object features:

3D morphological features:

• Volume: Object volume in µm³
• Surface Area: Outer surface area in µm²
• Equivalent Diameter: Diameter of sphere with same volume
• Sphericity: How sphere-like the object is (1 = perfect sphere)
• Width, Height, Depth: Bounding box dimensions in µm
• Center X, Y, Z: 3D centroid coordinates

3D intensity features:

• Mean, min, max, sum, standard deviation
• Measured in all intensity channels

Each row represents one 3D object in one volume.

Accumulate Records

Accumulate Records collects object counts across all volumes:

• Tracks count evolution over time
• Enables temporal analysis

Visualization

Object table (current volume): Shows features of all objects in the currently displayed volume.

Count table (accumulated): Displays object counts and statistics for all volumes.

Line chart: Plots object count vs. time, showing population dynamics.

Histogram: Displays distribution of selected feature (e.g., volume, sphericity) for current volume objects.

Results

This workflow produces:

• Per-volume object counts over time
• Individual 3D object measurements (volume, shape, intensity)
• Temporal trends in object population
• Feature distributions

Use cases: Organoid counting, spheroid growth analysis, 3D spot quantification, vesicle counting in volumes

Workflow 2: 3D particle tracking

Tracks particles moving in 3D space, calculating motion features including 3D trajectories and velocities.

  
---
config:
  look: handDrawn
  theme: neutral
---
graph TD
  subgraph Tracking
    time3d[Time and Center 3D]
    track3d[Track Particles 3D]
    accum[Accumulate Tracks]
  end

  subgraph "Object motion features"
    motion[Motion Features]
    posint[Position Integrate]
    table1[Table]
    scatter[Scatter Plot]
  end

  subgraph "Track features"
    trackfeat[Track Features]
    table2[Table - Tracks]
    scatter4d[Scatter Plot 4D]
  end

  subgraph "Presentation"
    layout1[Stacked Layout - Left]
    layout2[Stacked Layout - Right]
    display[Display]
  end

  subgraph Outputs
    savechn[Save Channels]
    savebin[Save Binaries]
    savetab[Save Tables]
  end

  %% Main workflow
  channels[Channels] -->|Cyan| spots3d[Bright Spots 3D] -->|Bin| time3d

  time3d -->|Records - Time, X, Y, Z| track3d
  track3d -->|Records - TrackId| accum

  accum -->|Records| motion
  accum -->|Records| trackfeat

  motion -->|Records - dX, dY, dZ| posint
  posint -->|Records| table1
  posint -->|Records| scatter

  trackfeat -->|Records - Tracks| table2
  trackfeat -->|Records| scatter4d

  %% Presentation
  table1 --> layout1
  scatter --> layout1
  table2 --> layout2
  scatter4d --> layout2

  layout1 --> display
  layout2 --> display

  %% Outputs
  channels -->|All| savechn
  spots3d -->|Bin| savebin
  display -->|Results| savetab

Time and Center 3D

Time and Center 3D extracts 3D positions:

Measured features:

• Time: Timestamp
• CenterX, CenterY, CenterZ: 3D centroid coordinates

Provides the position data needed for 3D tracking.

Track Particles 3D

Track Particles 3D connects particles across volumes in 3D space:

Parameters:

• maxSpeed: 146 µm/volume - maximum 3D displacement between volumes
• maxGap: 2 volumes - allows temporary disappearance

Algorithm: Calculates 3D Euclidean distance between particle positions:

distance = sqrt((x₂-x₁)² + (y₂-y₁)² + (z₂-z₁)²)

Particles within maxSpeed distance are candidates for linking. The algorithm chooses the globally optimal assignment.

Accumulate Tracks 3D

Accumulate Tracks collects tracks with:

• MinSegmentCount: 5 volumes minimum

Filters out short, unreliable 3D tracks.

Motion Features 3D

Motion Features calculates 3D motion parameters:

3D motion features:

• Speed: 3D velocity magnitude (µm/s)
• Direction: 3D heading (azimuth and elevation angles)
• dPositionX, dPositionY, dPositionZ: Displacement components
• Acceleration: Rate of 3D speed change

These capture full 3D motion dynamics.

Track Features 3D

Track Features summarizes entire 3D trajectories:

Per-track features:

• Length: Total 3D path length traveled
• Duration: Time span
• Speed: Mean 3D speed
• LineLength: Straight-line 3D distance start to end
• Straightness: LineLength / Length

Position Integration

Sequential Position Integrate reconstructs 3D trajectories by integrating displacement vectors:

Starting from an arbitrary origin, adds each displacement:

Position(t) = Position(t-1) + [dX, dY, dZ]

This creates smooth 3D trajectories suitable for visualization.

3D trajectory visualization

XY Scatter Plot: Projects 3D trajectories onto the XY plane, color-coded by track. Shows spatial distribution and horizontal movement patterns.

Scatter Plot 4D: Displays tracks in multidimensional feature space (e.g., speed vs. length vs. straightness), enabling population analysis.

Tables: Show per-volume positions and per-track summaries.

Results

This workflow produces:

• 3D track IDs linking particles across volumes
• Complete 3D trajectories (X, Y, Z, time)
• 3D motion features (speed, direction, acceleration in 3D)
• Track statistics (3D path length, 3D displacement)

Use cases: Vesicle trafficking in 3D, organelle movement, intracellular particle transport, cell migration in 3D matrices

Workflow 3: 3D object measurement

Measures morphology and intensity of 3D objects like cells or organoids, often combined with time-series analysis.

  
---
config:
  look: handDrawn
  theme: neutral
---
graph TD
  subgraph "Object features"
    meas3d[Object 3D Measurement]
    table1[Table]
    histo[Histogram]
  end

  subgraph "Object count"
    count3d[Object 3D Count]
    accum[Accumulate Records]
    table2[Table]
    chart[Line Chart]
  end

  subgraph "Presentation"
    layout1[Horizontal Layout - Top]
    layout2[Horizontal Layout - Bottom]
    display[Display]
  end

  subgraph Outputs
    savechn[Save Channels]
    savebin[Save Binaries]
    savetab[Save Tables]
  end

  %% Main workflow
  channels[Channels] -->|TRITC| cellpose3d[Cellpose 3D] -->|Cell| filter3d[Filter Objects 3D]
  filter3d -->|Cell| count3d
  filter3d -->|Cell| meas3d
  channels -->|All| count3d
  channels -->|All| meas3d

  count3d -->|Records| accum
  accum -->|Records| table2
  accum -->|Records| chart

  meas3d -->|Records| table1
  meas3d -->|Records| histo

  %% Presentation
  table1 --> layout1
  histo --> layout1
  table2 --> layout2
  chart --> layout2

  layout1 --> display
  layout2 --> display

  %% Outputs
  channels -->|All| savechn
  filter3d -->|Cell| savebin
  display -->|Results| savetab

Cellpose 3D segmentation

Cellpose 3D provides AI-powered 3D cell segmentation:

Parameters:

• Diameter: 60 pixels - typical cell diameter
• Anisotropy: 2 - cell model type

Cellpose 3D considers all Z-slices simultaneously, producing accurate 3D cell segmentations.

Filter Objects 3D

Filter Objects 3D removes objects by 3D criteria:

Filtering:

• Feature: Equivalent Diameter (3D)
• Comparator: Greater than
• Value: 5 µm

Removes objects smaller than 5 µm equivalent diameter, filtering out debris and segmentation artifacts.

Object 3D Count

Same as Workflow 1: provides per-volume summary statistics including object count, total volume, and mean intensities across volumes.

Object 3D Measurement

Measures comprehensive 3D features for each segmented object:

3D morphology:

• Volume, surface area
• Width, height, depth (bounding box)
• Sphericity, elongation ratios
• Position (center X, Y, Z)

3D intensity:

• Mean, max, min, sum across all channels
• Standard deviation within object

Each row represents one 3D object.

Accumulate Records

Collects per-volume counts across the time series:

• Tracks object count changes
• Accumulates summary statistics

Visualization

Per-object table (current volume): Shows 3D features of all objects in the displayed volume. Selecting rows highlights objects in the 3D viewer.

Per-volume table (accumulated): Displays object counts and aggregated features across all volumes.

Histogram: Shows distribution of any 3D feature (e.g., volume distribution, sphericity distribution) for current volume. Includes normal fit overlay.

Line chart: Plots temporal evolution of object count or aggregated features (e.g., mean volume over time).

Horizontal layouts: Organize views for easy comparison between current volume details and time-series trends.

Results

This workflow produces:

• 3D segmentation masks for all cells/objects
• Comprehensive 3D morphological measurements per object
• Intensity statistics in 3D volumes
• Temporal trends in object count and features
• Feature distributions with statistical fits

Use cases: Organoid growth tracking, spheroid morphology analysis, 3D cell shape characterization, volumetric tumor analysis

Comparing 3D workflows

Workflow selection guide

Parameter comparison

3D-specific challenges

Resolution anisotropy

Problem: Z-resolution typically 3-5× worse than XY resolution, causing:

• Distorted object shapes (spheres appear elongated)
• Biased measurements (overestimated surface area)
• Poor segmentation along Z

Solutions:

• Use isotropic acquisition when possible
• Set proper Z-scale in Cellpose 3D
• Apply resolution-aware measurements
• Resample to isotropic voxels if needed

Computational demands

3D processing requires significantly more:

• Memory: 100× more voxels than single slice
• Processing time: Volume operations are slower
• Storage: Larger file sizes

Solutions:

• Use binning (2×2 or 2×2×2) if resolution permits
• Crop to region of interest in XY and Z
• Process subsets of time points
• Use GPU-accelerated nodes when available

Z-drift and focus issues

Problem: Focus may drift during acquisition:

• Objects move out of focus
• Inconsistent segmentation across volumes
• Tracking failures

Solutions:

• Use hardware autofocus during acquisition
• Apply focus correction before analysis
• Crop Z-range to consistently focused region
• Use perfect focus systems

Limited Z-coverage

Problem: Objects may exit the imaged volume:

• Incomplete 3D structures at top/bottom
• Tracking breaks when particles leave volume
• Biased measurements near boundaries

Solutions:

• Acquire larger Z-range
• Filter border objects in Z dimension
• Account for partial volumes in analysis
• Use Remove Border Objects 3D node

Best practices for 3D analysis

Acquisition optimization

Z-stack parameters:

• Z-step: Follow Nyquist criterion: step ≤ λ/(2·NA·n) for optimal sampling
• Z-range: Cover full object extent plus margin
• Speed: Balance temporal resolution vs. photobleaching
• Focus: Use autofocus or perfect focus system

Resolution:

• Aim for isotropic if possible (same XY and Z resolution)
• At minimum: Z ≤ 3× XY spacing
• For sphericity measurements: Z ≤ 2× XY preferred

Signal quality:

• Adequate signal-to-noise ratio throughout Z-stack
• Minimize photobleaching (optimize laser power, dwell time)
• Uniform illumination across Z-range

Segmentation strategies

For sparse objects (particles, spots):

• Use Bright Spots 3D with appropriate diameter
• Set contrast threshold conservatively
• Verify detection at multiple Z-positions

For dense objects (cells, organoids):

• Use Cellpose 3D for best results
• Set diameter to typical object size
• Adjust Z-scale for anisotropic resolution
• Filter by size to remove artifacts

For irregular structures:

• Use 3D Threshold with appropriate cleaning
• Apply 3D smoothing before segmentation
• Consider watershed-based separation

Measurement validation

Visual verification:

• Inspect segmentation overlays in image and volume view
• Verify object boundaries at top and bottom of stack
• Check that 3D renderings match expected morphology

Quality metrics:

• Monitor segmentation consistency across volumes
• Validate against manual measurements

Advanced 3D topics

3D deconvolution

For improved resolution and contrast:

• Apply before segmentation
• Use appropriate PSF for your system
• Balance deconvolution iterations vs. artifacts
• See Deconvolution workflow

Surface rendering

For publication-quality visualization:

• Use 3D viewer with volume rendering
• Adjust transfer functions for optimal display
• Create rotating animations
• Export high-resolution images

Multi-channel 3D analysis

Analyze relationships in 3D:

• Segment each channel independently
• Calculate 3D colocalization
• Measure inter-object distances in 3D
• Analyze 3D spatial distributions

4D analysis (3D + time)

For dynamic processes:

• Track objects in 3D across time
• Measure morphology changes in 4D
• Analyze 3D cell divisions
• Track organoid growth over days

Computational requirements

Memory estimates

Processing time estimates

Relative to 2D processing:

• 3D Spot Detection: 10-20× slower
• Cellpose 3D: 20-50× slower (GPU recommended)
• 3D Measurements: 5-10× slower
• 3D Tracking: 2-5× slower (fewer time points usually)

Optimization strategies

Reduce data size:

• Crop to ROI before processing
• Bin 2×2 (reduces to 1/4 data)
• Process subset of time points first

Use GPU acceleration:

• Cellpose 3D benefits significantly
• Some measurement nodes support GPU
• Check node documentation

Batch processing:

• Process multiple files overnight
• Use Job Manager for automation
• Parallelize independent analyses

Troubleshooting 3D workflows

Poor segmentation

Symptoms: Objects split across slices, fragmented structures, missing objects

Causes:

• Inadequate Z-resolution
• Poor signal-to-noise ratio
• Wrong segmentation parameters

Solutions:

• Improve acquisition (smaller Z-step, better SNR)
• Use 3D smoothing before segmentation
• Adjust diameter and contrast settings
• Try Cellpose 3D instead of threshold

Incorrect measurements

Symptoms: Unrealistic volumes, wrong sphericity, inconsistent values

Causes:

• Incorrect calibration (especially Z)
• Anisotropic resolution not accounted for
• Border effects

Solutions:

• Verify XY and Z calibration
• Check metadata in image file
• Set proper Z-scale in analysis nodes
• Remove border objects

Tracking failures in 3D

Symptoms: Broken tracks, ID swapping, missing particles

Causes:

• MaxSpeed too small for 3D movement
• Particles leaving Z-range
• Segmentation inconsistency

Solutions:

• Increase maxSpeed to account for Z-displacement
• Use maxGap to bridge temporary disappearances
• Improve segmentation consistency
• Crop to central Z-region

Performance issues

Symptoms: Out of memory errors, excessive processing time

Causes:

• Large 3D datasets
• Insufficient RAM
• Inefficient processing

Solutions:

• Process smaller Z-ranges
• Use binning to reduce data size
• Add more RAM
• Process volumes in batches

Related workflows

• Object counting - 2D object counting baseline
• 2D tracking - 2D tracking methods
• Cell measurement - 2D cell analysis
• Deconvolution - 3D image enhancement